Autism is a severely handicapping condition with onset in early childhood. Most persons with autism require lifelong supervised or institutional care. They exhibit deviant and/or delayed development in intellectual, social, emotional, and self-help skills, often accompanied by high rates of aggression, self injurious, and ritualistic behaviors. Several recent studies have reported favorable outcome of early and intensive behavioral intervention in autism. Of these studies, the Lovaas (1987) study has had the most comprehensive and long term follow-up (McEachin, Smith, & Lovaas, 1993). This study also contained a number of methodological safeguards. Follow-up data show that an intensively treated group (n=19) of preschool aged children with autism achieved major gains in intellectual and academic functioning, in contrast to a similarly constituted control group who did not evidence such improvement. The proposed research will test the replicability across three sites of the favorable treatment outcome in the Lovaas (1987) study. The three sites (University of Oslo, Bancroft Schools, and the Alta California Regional Center) have each submitted letters of intent to collaborate. The sites are well prepared and have secured adequate referrals of subjects and financial support. Supervisory staff from sites will be trained in treatment and assessment at the UCLA site. Strict quality control measures will assure that assessment and treatment are uniform across sites. The proposed research employs an experimental design and methodological safeguards. Ninety subjects (30 from each site) will be randomly assigned into either an intensively treated experimental group or a parent training control group, each with 45 subjects. The experimental group will receive 35 hours a week of 1:1 behavioral treatment in their homes and community for 2 years. The control group will receive parent training for 6 months. Subjects will be between 24 - 42 months of age at intake, have a diagnosis of autism, and obtain a ratio IQ above 35. A large number of normed psychological tests will assess intellectual, academic, social, emotional, and self help skills before, during, and after treatment. Three hypotheses are tested. First, independent sites can reliably and validly replicate the diagnostic, assessment, and treatment procedures used at the UCLA sites. Second, independent sites can produce significant increases in intellectual, academic, language, adaptive, and socioemotional functioning after intensive treatment. Third, subjects who achieve the best outcome in the experimental group can be identified based on pretreatment assessment.